3D Visualization for Matlab

matVTK

3D Visualization for Matlab

Erich Birngruber, René Donner, Georg Langs

Computational Image Analysis and Radiology LabMedical University of Vienna, Austria

http://www.cir.meduniwien.ac.at/

CSAIL, Massachusetts Institute of Technology, USA

Dienstag, 29. September 2009

Motivation

2

Previous work

• Gobbi et al. 2008: Simulink Libraries for Visual Programming of VTK and ITK

Need

• Work on (large) medical datasets

• Need for “visual debugging” / big picture

• Easy usability for the Matlab userExample from work of S. Witoszynskyj

Dienstag, 29. September 2009

Motivation

Matlab

• Rapid prototyping / algorithm development

• Limited visualization for 3D data

3

Dienstag, 29. September 2009

Motivation

Matlab

• Rapid prototyping / algorithm development

• Limited visualization for 3D data

VTK

• Powerful, good performance

• Platform independent

3

Dienstag, 29. September 2009

Outline

• Development foundations

• Resulting Framework

• Use cases

Dienstag, 29. September 2009

Foundations

• Mex interface allows access to C/C++ libs

- Use Matlab memory managment

- Mex entry point function

void mexFunction(int nlhs,mxArray *plhs[],int nrhs,const mxArray *prhs[]){ ... }

• VTK “pipes & filters” concatenate data processing

5

Dienstag, 29. September 2009

Obstacles

• Keeping state:

no direct support for handles in Mex

• VTK rendering loop interferes with Matlab

6

Dienstag, 29. September 2009

matVTK Framework

• Handle based:

implicit creation, automatic destruction

• Imitates Matlab plot commands:

prefixed with vtk...

vtkplotpoints(pointset)

vtkquiver(points, vectors)

7

Dienstag, 29. September 2009

Parameters

• No code generation: explicit mapping

• Easily extendable

• Basic settings mapping:

8

Dienstag, 29. September 2009

Parameters

• No code generation: explicit mapping

• Easily extendable

• Basic settings mapping:

VTKscalar valuetriplet valuedifferent modes (Constants)

8

Dienstag, 29. September 2009

Parameters

• No code generation: explicit mapping

• Easily extendable

• Basic settings mapping:

VTKscalar valuetriplet valuedifferent modes (Constants)

mapping

8

Dienstag, 29. September 2009

Parameters

• No code generation: explicit mapping

• Easily extendable

• Basic settings mapping:

VTKscalar valuetriplet valuedifferent modes (Constants)

Matlab5[1, 2, 3]“String”

mapping

8

Dienstag, 29. September 2009

matVTK Framework

9

Matlab

VTK

Input

Data

Numeric

Output

Handle

Graphic primitives

points

lines

surfaces

volumes

Algorithms

triangulation

pointselection

Visual

output

Function

callsHandle

Dienstag, 29. September 2009

Primitives

10Example from work of E. Dittrich

Dienstag, 29. September 2009

Primitives

• Points

10Example from work of E. Dittrich

Dienstag, 29. September 2009

Primitives

• Points

10

• Lines

Example from work of E. Dittrich

Dienstag, 29. September 2009

Primitives

• Points

10

• Lines

• Volumes

• Meshes (surfaces)

Example from work of E. Dittrich

Dienstag, 29. September 2009

Primitives

• Points

10

• Lines

• Volumes

• Meshes (surfaces)

• Vectors (as arrows)

Example from work of E. Dittrich

Dienstag, 29. September 2009

Primitives

• Points

10

• Lines

• Volumes

• Meshes (surfaces)

• Vectors (as arrows)

• Tensors

Example from work of E. Dittrich

Dienstag, 29. September 2009

Combining Primitives

• Combinations of primitives can construct complex scenes

11

Example from work of E. Schwartz

Dienstag, 29. September 2009

Additional Functionality

• Widgets for data inspection & annotation

• Cropping

• Labels

• Axes

• Data export

• Restore scenes

12

Dienstag, 29. September 2009

Example

% plot and save components to handlehv = vtkplotvolume(volume,‘SoftSkin’)hm = vtkplotmesh(vertices, faces ‘color’, [0 0 1])% crop volume and mesh with plane widgetvtkcrop(hv, ‘plane’)vtkcrop(hm)% show axesvtkgrid% interactive display, free resourcesvtkshowvtkdestroy

13

Dienstag, 29. September 2009

Example

% plot and save components to handlehv = vtkplotvolume(volume,‘SoftSkin’)hm = vtkplotmesh(vertices, faces ‘color’, [0 0 1])% crop volume and mesh with plane widgetvtkcrop(hv, ‘plane’)vtkcrop(hm)% show axesvtkgrid% interactive display, free resourcesvtkshowvtkdestroy

13

Dienstag, 29. September 2009

Performance

100 130 160 190 220 2500

10

20

30

40

50

60

70

cube sidelength

se

cs

matVTK

Matlab

Isosurfacing for different volume sizes14

Dienstag, 29. September 2009

Availability

• Developed on MacOS X 10.5

• Linux 32bit, 64bit

• Windows XP, Vista, 7

15

as platform independent as Matlab and VTK

Dienstag, 29. September 2009

Outlook

• Streaming visualization

• Saving widget states

• Remove scene objects

• Event recording / movies

16

Example from work of G. Langs

Dienstag, 29. September 2009

Thanks

• Available online:

http://www.cir.meduniwien.ac.at/matvtk/

• Feedback and ideas welcomeerich.birngruber@meduniwien.ac.at

17

Dienstag, 29. September 2009